Biological Foundations?

[Sergio Navega]

Detlef Morgenstern wrote:
> Some time ago I asked myself this question:
> What if my boss came and said: "Design a brain!" ?

That's indeed two different questions in one. Your boss may want
you to design a brain sharing the fundamental operating principles,
which is a bunch of simple elements massively interconnected (neural
networks). Or he may want just the functionality (the "end result").
For this purpose, a monoprocessor von Neumann machine would suffice.
There is AI research today focusing both paradigms, although the
latter is accumulating frustrations for some time.

In any case, the question of information compression is
significant. And that's because our brains are machines of modeling
the world, reducing the complexity of the events to manageable
(and "memorizable") chunks.

> I would try to make the processing unit as primitive as possible (you see
> some related considerations in my other postings). Which is the absolute
> minimum? Evolution's brain seems to have still too much complexity in its
> processing units (neurons), they seem to have been in a hurry when
designing
> this. And now they must switch connections.

The architecture of such a "machine" may seem important at first. But I
believe it will just influence performance. What is necessary is obeying
the basic principles. It is in defining what these principles are that
dissonant voices are heard among the AI community.

For me, the most basic principle is inductive generalization, the ability
one intelligent entity has of perceiving regularities in the world and
grouping them into cohesive structures. This can be seen as information
compression, although I think compression happens by "accident". What
the brain (and future AI machines) pursue is the transformation of
patterns of regular occurrences into abstract mental entities that
can be manipulated. This is what allows our brain to model the world
and experiment with it "internally". The "coincidence" of obtaining
compression when executing this process of induction is interesting:
it may indicate something important in our quest to understand
intelligence.

Regards,
Sergio Navega.